Customers welcome GenAI in service, as long as a human is one click away

Generative AI is now a regular part of how people reach the companies they buy from. The open question is no longer whether to use it. It is how to use it without pushing customers away. A recent Gartner survey gives a clear answer, and it is one worth reading closely if you own the customer relationship.

The headline finding: 87% want a way through to a human

Gartner surveyed 3,566 B2B and B2C customers in February and March 2026. The standout number: 87% say it is essential that a company using GenAI for service still offers a way to reach a human agent. Nearly nine in ten customers, in other words, treat human access as a condition, not a nice-to-have.

That does not mean customers are rejecting the technology. Half of them (50%) say their interactions are actually easier when a company uses GenAI. The picture is not "AI versus people". It is customers who are open to AI, on the condition that it never becomes the wall between them and a person who can help.

Why the split matters more than it looks

The two findings sit side by side because customer opinion has been divided for a long time. As Gartner's Eric Keller put it to CX Dive, roughly half of customers have leaned towards self-service and efficiency for years, while the other half have wanted to talk to a person. Neither group is going away.

So the design goal is not to pick a side. It is to serve both from the same experience: fast, capable AI for the customer who wants a quick resolution, and a clear, easy route to a human for the customer who wants one, or who started with AI and hit its limit.

The cost of getting this wrong is concrete. When customers are forced through several failed AI exchanges before they can reach a person, they are less likely to come back to that tool at all. The shortcut to containment becomes the reason adoption stalls.

What "good" looks like

Keller sets out three principles for GenAI in service that customers actually accept. The AI should be transparent that it is AI, it should explain its answer and how it reached it, and it should offer a clear off-ramp to a human.

The transparency point carries real weight. When AI pretends to be a person, it erodes trust the moment the customer works it out. Being upfront costs nothing and protects the relationship.

The off-ramp is where many companies hesitate, because a route to a human looks like a route around their efficiency targets. The evidence points the other way. Make it easy to switch to a person, and customers use the AI more over time, because they know it is a safe place to start and that help is close if they need it. The easy exit is what makes the front door worth using.

The handoff is the moment that counts

If there is one detail that decides whether a customer leaves satisfied or frustrated, it is the transition from AI to human. Done badly, the customer repeats everything they have already said, and the earlier AI exchange becomes wasted effort that reads as indifference.

Done well, the human agent picks up with full context: what the customer came in for, what the AI already tried, where it got stuck. The customer moves forward instead of starting over. That continuity is not a technical nicety. It is the difference between a service experience that respects the customer's time and one that spends it.

What customers now expect AI to do

There is a second shift in the data that reshapes the brief. Customers no longer come to GenAI only for answers. 58% of those who use GenAI have used it to complete a task on their behalf, and in B2B that rises to 74%.

Booking an appointment, placing an order, submitting documents, managing a subscription, escalating a request: these are actions, not questions. The implication for service design is direct. The most useful pattern pairs conversational AI that understands what the customer wants with automation that completes the transaction in the background. AI reads the intent. The system does the work. The customer gets it done.

Worth noting too: customers increasingly begin outside company channels entirely. In their most recent service interaction, they were around three times more likely to reach for a third-party GenAI tool like ChatGPT, Gemini or Copilot than a company-provided chatbot. People still come back to the company when they need to transact, reach account-specific information or resolve something complex. But the first step often happens somewhere you do not control, which raises the bar for the experience you do.

What this means for how you design service

Pull the threads together and the direction is clear:

• Do not make GenAI a mandatory gate. It works best when it gathers information, understands intent and attempts a resolution only when confidence is high, with a visible path to a person throughout.
• Be honest that it is AI. Transparency protects the trust everything else depends on.
• Engineer the handoff. Carry the full context across so no one repeats themselves.
• Design for action, not just answers. Let AI understand the request and let automation complete it.
• Meet customers where they start. Some arrive through third-party AI. The experience you own has to be good enough to bring them home.

None of this is a technology problem waiting for the next model release. It is a relationship problem, and it is answered by understanding what your customers want at the moment they reach out: a quick resolution when AI can give it, and a person when it cannot, with no wall in between.

Get that balance right and AI stops being a cost-containment tactic customers resent. It becomes part of why they stay.

Source: Gartner, "Gartner Survey Finds 87% of Customers Say Companies Using GenAI for Customer Service Must Provide Access to a Human Agent" (4 August 2026)